ChatGPT is plugged into a lot of apps and more so with every passing day. People are plugging into ChatGPT in record numbers. Then OpenAI creates customizable micro GPTs and a store to sell them in. Following that announcement came the debut of a mobile, wearable AI pin by Humane. The AI pin is priced at a modest $699 which is substantially cheaper than many modern smartphones. Traditional apps and websites are squeezed somewhere in the middle of these rapid-fire developments.
ChatGPT, like other generative AI (GenAI) models and applications, are adept at understanding and delivering what people say they want. Given that consumers and businesses alike can summon most of what they need from GenAI via a simple voice, image or text prompt, what reason would they have to go directly to individual apps and websites?
Will this lead to the end of apps as we know them? And will this kill peripheral industries like search engine optimization (SEO) and online advertising? Perhaps we’ll wait and see or maybe poke around and find out. Either way, it’s a fairly safe bet that a shakeup lies ahead.
GenAI in the Super App Role
A “super app” simply means a single application that can handle any and all requests. If you’re picturing HAL from the movie “2001: A Space Odyssey,” you’ve got the gist of a super app, even though your mind turned to a super dark example of a type of AI that does not exist. At least not yet.
But it isn’t incorrect to picture a super app as an AI model with a single, unified user interface. Its ideal form would be an artificial general intelligence model befitting sci-fi lore, although the generative AI models that exist today are powerful enough to be the next best thing.
“A super app necessitates human level social intelligence, which we haven’t even scratched the surface of. If we are thinking of J.A.R.V.I.S. from Ironman, its social skills are such a critical part of its communication,” says Rahul Rajan, an expert in human-computer interaction and the co-founder and chief AI officer at Uplift Labs.
GenAI and its applications like ChatGPT are charming or conversational in most but not all encounters with humans. Unfortunately, when it does present charming conversation it can be difficult to tell if it’s being helpful or not.
There are solid reasons not to trust AI. ChatGPT and other generative AI and large language models (LLMs) have proven to be unreliable as they suffer frequent bouts of hallucinations. They can outright lie too, even when there’s no need to misrepresent the data.
“For example, you could ask a conversational model what color the sky is, and it will answer “blue” as most of the text it has processed says so, but it will eventually answer “red” or “yellow,” says Jose Selvi, executive principal security consultant at NCC Group.
This problem is particularly common in AI applications like ChatGPT when it’s used in a one-size-fits-all-problems approach. Meaning ChatGPT is used to answer any question, anytime, ad infinitum. Micro GPTs are less prone to this issue because they are specialized and typically deal with one issue. Even so, GenAI of all stripes are buggy.
That’s a problem when accurate information is vital — such as in healthcare, a legal issue, or any number of queries where life, limb, liberty, money, or simple peace of mind are at stake.
“To cede control [to AI], you need to trust, and to trust tests a very primal innate human instinct. It’s all possible of course but it’s less about the technical prowess of AI and more about the fuzzy areas of what makes us human,” Rajan says.
Phoning It In
But there’s more at work here than a shortage of social niceties and trust. It turns out that glib conversation isn’t the path of least resistance in completing some tasks.
“Many user interfaces (UIs) are still more efficient than voice or text [AI prompts]. For example, it is faster for me to click some buttons to crop a photo than it is to specify the precise dimensions,” says Randall Hunt, vice president of Cloud Strategy and Solutions at Caylent.
If using apps is — or is perceived to be — easier than using AI, users will likely continue to use them, at least in the short term. “Long term maybe all of these interface issues are resolved by increasingly advanced technology,” Hunt says. Interestingly, some of those more advanced technologies, like autonomous AI agents, are already here.
Then there’s the mobile, wearable AI pin recently introduced by Humane. It’s reasonably priced below many modern smartphones and may be the harbinger of the end of phone devices. Here’s a short video on how it works.
The AI pin, and no doubt similar mobile AI form factors from glasses to tattoos, will phone in our work for us — sans phone.
Channel This
There are those who say that asking if AI will eat apps is the wrong question. It may be more of a matter of individual preferences in communication channels. At least in the short term.
“There are websites, mobile apps, and also call centers, and all of them are interfaces between customers and the company. AI chatbots will be an additional interface that we will soon be seen in many companies, but people still will be able to choose which interface they prefer, in the same way some people prefer using the website and others calling the call center, for example,” says Jose Selvi, executive principal security consultant at NCC Group, a global security consultancy.
There are others who say that envisioning AI models as simply another channel is far too narrow and also off base.
“One way of thinking about generative AI is as a new and powerful operating system. Like a traditional operating system, generative AI provides extremely general resources and tools –sometimes too general,” says Michael Kearns, professor of computer and information science at the University of Pennsylvania’s School of Engineering and Applied Science.
“So I envision a future where apps and websites still exist, but are powered by an underlying generative AI operating system. And for less task-specific, open-ended interactions, the current playground interfaces will still be valuable and fascinating to users,” Kearns adds.
App-plosion vs App-ocalypse
There is a camp that thinks AI will cause an explosion of apps — not their demise. Granted that some, but not all, members in this camp are developers who make their living creating and selling traditional apps. But that doesn’t mean they’re wrong.
“AI will not kill apps, we may in fact see the exact opposite — a massive uptick in apps as the core business logic, algorithms, and trade secrets around workflow become less complex. All someone will need to do is build a front end and create interfaces for use cases,” says Jimmie Lee, CTO at XFactor.io, an AI-powered go-to-market platform.
“Ultimately, this provides simplicity as most of the population is not tech-savvy enough to understand the intricacies of generative AI and will favor the integration of tools like ChatGPT via buttons or widgets in apps,” Lee adds.
However, that’s pretty much what OpenAI is doing with micro GPTs and its soon to be launched GPT store. Meaning AI may have beat developers to the punch by flipping the model to AI with apps underneath, rather than apps with AI stuffed in a widget.
Even so, we’ll need to wait to see which is victorious in the end. There remains a considerable number of wild cards in play for now.
“Generative AI models can and should be used to support and amplify apps, not replace them. AI-powered apps can provide users with the ability to process information faster, connect unconnected data to build richer context and hence make quick, data-driven decisions, which improves their in-app experience,” says Sunil Senan, Infosys SVP and global head of data, analytics, and AI.
In the end, whether apps or AI win this battle comes down to which provides more value.
“Simply adding chatbots to apps and services because competitors are doing so won’t cut it anymore, and organizations are realizing their AI-backed apps need to add more value to users,” Senan adds.
